Latest / The Marketing Meeting with Itir Eraslan / 36. Transforming Marketing Workflows with AI and Automation with Reid Robinson
Transcript
- 0:00Reid Robinson: Within AI, the thing that becomes critical, I would say,
- 0:02is getting these models context.
- 0:05Typically, in the workflows that are valuable, you're not running things
- 0:08where you're just blanket asking the model to give you something.
- 0:12You know, you're not just saying, generate a blog article about automation.
- 0:15That's not going to get you a high value output, right?
- 0:18Like what can give you information that, when provided to an LLM, will
- 0:23give you an output that feels valuable.
- 0:25Itir Eraslan: Hi, this is the marketing meeting and I'm your host Itir Eraslan.
- 0:30Every two weeks I meet with experts and we talk about topics related to
- 0:34brands, marketing, and businesses.
- 0:36We sometimes add random lifestyle topics too.
- 0:39I hope you enjoy the show.
- 0:45Welcome to the marketing meeting podcast.
- 0:47I'm happy to host Reid Robinson today.
- 0:50He's the lead product manager at Zephyr.
- 0:53Zephyr is a workflow automation software company as A lot of marketers would know.
- 0:59Uh, great to meet you, Reid.
- 1:01Welcome.
- 1:02Reid Robinson: Thanks.
- 1:02Thanks for having me on.
- 1:03And hello, everybody.
- 1:05Itir Eraslan: Reid, your LinkedIn headline has three words, which
- 1:09it says building AI products.
- 1:12That's the, one of the main reasons that I wanted to really talk to
- 1:16you because I want to talk to.
- 1:19the expert.
- 1:21But there is also another thing in your profile.
- 1:24It says you are a student of social media.
- 1:27So it gives a clue about your personality and your style.
- 1:32Because instead of saying you're an expert of social media, you are saying
- 1:36you're a student of social media, which is quite important for me.
- 1:39Because when I look to your background, you worked at HootSuite, now at
- 1:44Zapier, and your And these are very important platforms for marketers.
- 1:49That's why, I mean, like, I would like to ask questions about automation,
- 1:53especially, which is a big question mark, uh, especially on marketer's side.
- 1:59After some announcement say that 95 percent of, uh, marketing
- 2:03agencies will disappear.
- 2:05So it's all about automation and what will automate soon, what will
- 2:08not, and those type of questions.
- 2:11So I would like to start with a broader question, and then we can deep dive
- 2:15to some of the practical things that can help listeners on this journey.
- 2:20So which areas of marketing do you think AI will have the most significant
- 2:26impact in the, let's say, next one year?
- 2:31Reid Robinson: The next one year for marketers, I think, I want to say
- 2:34personalization of some of the things.
- 2:37Um, what I mean by that is I think some of the ones we've seen really,
- 2:39I think impact today, a lot of users doing things like, uh, content draft,
- 2:44uh, and stuff like that, but I don't know if draft content is going to have
- 2:48like a massive impact for marketers.
- 2:51Uh, I think one of the things from, at least for a marketing use case
- 2:54is going to be around, uh, being able to personalize some customer
- 2:58communication in ways that, like, as you.
- 3:01Pointed out, I've worked in the world of like marketing tech for quite a while.
- 3:05And I feel like along that journey, personalization has been like a
- 3:09hot topic for a very long time.
- 3:12And I actually think, you know, done it through various ways of, okay,
- 3:15we're personalizing the links in this email template, or we're personalizing
- 3:19the subject line of these emails.
- 3:22But with generative AI, you can really do.
- 3:25And an impressively wild amount of personalization, like you
- 3:29could do the entire email.
- 3:30And yeah, I think that's probably what I think I'm most excited about.
- 3:35One of the initiatives I helped internally, uh, was doing like an
- 3:39automated marketing campaign to recommend additional use cases that
- 3:42did use like a generative AI step to help, in essence, hallucinate an
- 3:47additional use case for that user.
- 3:50But we actually ended up seeing a lift in activations by recommending
- 3:55additional use cases through a generative AI step, which was cool.
- 3:59Um, and I think that'll become more and more accessible and
- 4:03more and more commonplace.
- 4:05Um, and I think as a user as well, it'd be nicer to start to see some comms come
- 4:09in that are like, Hey, we know you were tinkering with this and you tried this and
- 4:14like getting down to a much more personal level on how those comms get across.
- 4:20Um, it's gonna be really cool.
- 4:22I think of that more from an email side.
- 4:23I've yet to be clear on what that looks like from a social media side.
- 4:27Itir Eraslan: So you mentioned that for email personalization,
- 4:31uh, would be one of the key things.
- 4:33But you're not sure about social media now?
- 4:37Reid Robinson: I meant in terms of how you would get personalization across
- 4:39to an end user, other than advertising, which is kind of doesn't scale all that
- 4:43well to do a very niche thing, right?
- 4:45Like you can do a one to one personalization effort at scale
- 4:48with email very effectively, as well as in product, I should say.
- 4:52Right.
- 4:52Like you're in product marketing capabilities.
- 4:54Um, also, you know, especially as these models can get faster and faster,
- 4:58you can start to do things there.
- 5:00Itir Eraslan: I have one question, for example, you know, I'm, I'm a marketer and
- 5:03I've worked in various projects, sometimes it's a jewelry, high end jewelry brand.
- 5:07Sometimes it's a tech brand.
- 5:09Sometimes it's a kids furniture brand or so on.
- 5:13And then, you know, for these brands, I do.
- 5:15Search on platforms, you know, Google search and then now including chat,
- 5:20GPT, Instagram, social media, and so on.
- 5:24I mean, like you can understand by going through my social feeds, which type of
- 5:28project I'm doing, I'm either like, they are expecting me to have a baby soon
- 5:33because I'm checking kids furniture or, uh, they expect that I'm a tech, uh,
- 5:38SaaS marketer or something like that.
- 5:40It's really screws up.
- 5:42I'm a veno where I.
- 5:43Got into a project.
- 5:44It screws up my social media because we assume that, you know, this person is
- 5:49suddenly interested in for kids furniture.
- 5:52Uh, but it's, it's not the case.
- 5:53I mean, like, I think in that case, social media is a little bit
- 5:57trickier to understand, I guess.
- 6:01If you don't know the context to that person's personal life, I guess.
- 6:07Reid Robinson: Yeah.
- 6:09Uh, yeah, this one can get very deep in terms of how, yeah, the
- 6:12world of, of niche advertising and content and personalization.
- 6:17Um, and how much people want that or not.
- 6:19Um, actually, yeah, that honestly is a, is a fun tightrope walk.
- 6:23I'm sure all marketers are familiar with, particularly those on the
- 6:26ad for ties inside of things.
- 6:28Um, it's yeah, it's fascinating.
- 6:31Honestly, I probably don't have the greatest insights to say here, other
- 6:34than the fact that I really do hope for a future that does have like more
- 6:38transparency of what information you're like giving up from yourself to get
- 6:44more personalized recommendations.
- 6:46Um, I think that would be really cool to start to see in some search things of
- 6:50like, Hey, for this search, I do want you to understand my personal life, right?
- 6:55Whereas if I do another search, maybe that search is only my work information.
- 6:59Uh, it's using for context.
- 7:01Um, could be cool to start to get that level of like granular control when you
- 7:06are looking to help from an algorithm or an AI model or whatever it is.
- 7:13Itir Eraslan: Maybe, I mean, like one day since we can just start to customize,
- 7:16you know, it's AI saying that, okay, I'm now, I'm going to ask questions
- 7:20about my, uh, business and my strategy.
- 7:23And then it then includes when it's suggesting an Instagram feed, then
- 7:27it excludes those types of things.
- 7:29Uh, but, um, what are the areas since you are so much into automation with Zapier?
- 7:35Uh, what are the areas that marketing automation will be
- 7:38adapted fast and very soon?
- 7:41I know that a lot of marketers have already adapted them.
- 7:44Uh, but still, as I see from my friends and my colleagues.
- 7:49A big portion of them haven't adopted, uh, marketing automation yet.
- 7:54Reid Robinson: So the, the top use cases that we see, um, is users
- 8:00using AI to generate things like draft social media posts, blog,
- 8:05and just overall marketing content.
- 8:07And I'll give you an example of use cases would be things like you have a, an
- 8:13automation run where every time you add a.
- 8:16A work topic to like an air table, you know, if you put it to a certain status
- 8:20and air table or to a Trello card status, right, you can then kick off a kind of
- 8:25draft creation process where you might have given a something like a chat GPT
- 8:31step, the access to, you know, some sort of Information on how you typically
- 8:36write what your company does what your angle tries to be Might as well include
- 8:42like examples of past blog articles and to try to mimic the style that you're
- 8:47going for And then you might in like the actual run might include some context of,
- 8:54Hey, we want to do a campaign about X.
- 8:56This article should be for Y.
- 8:59Can you help generate a draft blog article?
- 9:01And then subsequent steps might try to convert that to a social media post and
- 9:06like for Twitter and the LinkedIn one, and then a Instagram recommendation, and
- 9:11then kind of output that to a marketer so that you have, you A lot of like,
- 9:16you're just at a closer state, I'd say at this point, definitely more of like a
- 9:20draft is the right way to think about it.
- 9:22The other side, they would have saw it as like converting, um, I'm sure
- 9:25marketers here are very familiar with the concept of like taking a long
- 9:28form article and then repurposing it in various different formats.
- 9:31Uh, the models also pretty good at getting you further along, right?
- 9:36Like, Hey, our team or a company that we are partners with just
- 9:40published this blog article.
- 9:41Can you help recommend like what our take on this should be?
- 9:44You're like, what could tweet for this be?
- 9:46I think back to the days at Hootsuite, we're like RSS feeds were a big thing
- 9:50where people were just reposting from an RSS feed and kind of like that.
- 9:55You can do something similar, but at least now you can find it.
- 9:58Start to personalize it and you can start to make it about your brand and you can
- 10:01start to make connections, um, and get it to a much closer state, um, than you
- 10:07would by just looking at a raw RSS feed.
- 10:10Um, you could even add some like, I'll put the word intelligence in air
- 10:14quotes, uh, for anybody listening, uh, to that flow of saying like,
- 10:17Hey, you know, this is the RSS feed.
- 10:20Can you tell me which of these articles do you think are like
- 10:22relevant to our business priorities?
- 10:24Right.
- 10:25And then so you can actually start to like triage your RSS feeds of content as well.
- 10:31Itir Eraslan: Uh,
- 10:31Reid Robinson: so things like that, we are probably, I want to be clear,
- 10:33like, number one, um, that we're seeing today, number two, uh, is interacting
- 10:39with customers and drafts for that.
- 10:41So whether you want to automate responses to customer inquiries and
- 10:44give it information on like your, Business, whether that's the sales
- 10:48or supporting queries, uh, reviews, helping you automatically reply to
- 10:53customer reviews, um, things like that seem to be very helpful after that data
- 10:59extraction, this is more on like the, if anybody on the side is like a lead
- 11:03generation side of marketing, that's all those times where you have like
- 11:06unstructured data that you just like kind of wish came to you as a form.
- 11:10Anytime you want that great use case.
- 11:13Um, so you can basically tell these models, Hey, You're going
- 11:16to get this sort of structure.
- 11:19Maybe it's like an email, but it's going to come in like a.
- 11:21All the text in one box, and I need you to help extract the subject line, the
- 11:26body, the first name, the last name, the company they work for, and like, give
- 11:30me values that then I can put to my CRM.
- 11:32It does that incredibly well.
- 11:34Itir Eraslan: Uh,
- 11:35Reid Robinson: Summarization is then a great one.
- 11:37So again, going back to that RSS feed style thing, like, You know, take
- 11:41all the articles published from, I don't know, fortune on a particular
- 11:45topic over this period of time.
- 11:47And, you know, you can have that thing running.
- 11:49And then at the end of the week, maybe you get like bullet points of which
- 11:53ones are relevant to your business and summaries of those articles.
- 11:56Right.
- 11:57Um, And then, yeah, like I said, that kind of like air quotes again,
- 12:01intelligence layer for recommendations.
- 12:04Um, so you can give it information about yourself, your business,
- 12:07your role, the objectives.
- 12:09And, you know, when given a bunch of other text, you can say like,
- 12:12hey, help me whittle things down.
- 12:15Help me figure out what is, um, strategy there.
- 12:19So that's what we're workflows, to be clear.
- 12:22But I think I'll touch on one other thing, which is like, Expanding your capabilities
- 12:27just as a, a worker or as an individual.
- 12:30Um, I think that's probably one of the coolest things for me.
- 12:33Uh, is the fact that you can leverage, you know, tools like chat to PT, Gemini,
- 12:37Claude, whatever, um, to just learn more and do more than you previously could do.
- 12:43Um, for marketers that might be learning how to use Zapier, right?
- 12:48Maybe in the past you've tried to use Zapier and it's confusing.
- 12:51Well, now you can like talk to chat GPT.
- 12:53I built a, an automation consultant GPT.
- 12:55That's actually been used I think 35, 000 times at this point, which is just
- 13:00helping people learn about like what they can do with automations and it
- 13:03recommends Zapier workflows for them.
- 13:06Um, And things like that get me really excited because you can learn coding,
- 13:12you can learn like all sorts of neat things that you weren't previously
- 13:16capable of doing that now you can do.
- 13:18Itir Eraslan: Just one question, when you think about like those, the last
- 13:21point is about improving yourself, uh, about your capabilities and then you ask
- 13:26how to automate and is ChatGPT coming up already with recommendations on Zapier or?
- 13:34If you ask for Zapier, that it comes with recommendations about Zapier.
- 13:39Reid Robinson: It depends.
- 13:40It sounds like if I'm understanding the question, there is this
- 13:43kind of question in the industry right now, like SEO for models,
- 13:48Itir Eraslan: um,
- 13:48Reid Robinson: right.
- 13:48Which is like without prompting, right.
- 13:52Without telling it explicitly, you're interested in Zapier.
- 13:55Is the model going to recommend Zapier, right?
- 13:57If I just say, Hey, I'm looking to automate my job.
- 14:01Can you give me ways to do that?
- 14:03I probably would recommend Zapier.
- 14:05Um, I don't really know, there's a fascinating field of marketing emerging
- 14:11for like SEO because then the following question is, how can we influence that?
- 14:16Right?
- 14:16Like, how can we as marketers influence?
- 14:18That's
- 14:18Itir Eraslan: why I was asking this question.
- 14:20Reid Robinson: Yeah, well, it's a fun topic because I mean, OpenAI announced
- 14:23this week that they are working on training their next big model.
- 14:28And, you know, with that comes scraping the entire internet, essentially.
- 14:31Right.
- 14:32And yeah, as a marketer, you're like, well, we have a lot
- 14:35of content on the internet.
- 14:36Are there things we can do that then help us rank higher in this like raw
- 14:40response from these foundational models?
- 14:43I wish I could tell you more aside from the fact that it's a very
- 14:46interesting unknown right now.
- 14:49Um, for me, I think more of like, you know, an actual
- 14:53custom version of GPT for GPTs.
- 14:57It's.
- 14:57told to focus on Zapier and we give it a tool that helps it like build
- 15:02workflows for users based on their own description of what they're looking to do.
- 15:07That sort of thing for now.
- 15:09Itir Eraslan: This is fascinating for me that because there are
- 15:11two things that fascinates and it makes me curious about this, which
- 15:14is the fact that I asked already.
- 15:16So would one day generative AI would be suggesting me.
- 15:21Uh, platforms, because right now, if you don't specifically ask for a platform,
- 15:26uh, name, like bring me names of the best automation platforms, it doesn't, it
- 15:32sounds quite objective, but I feel that, you know, OpenAI Uh, Gemini, Google, uh,
- 15:40they're going to soon, uh, turn it into, I'm just waiting for the day when they're
- 15:44going to announce, okay, we are going to take, start taking ads and we are going
- 15:47to start, you know, it's not going to be like only an organic thing and so on.
- 15:52It's going to be done.
- 15:54Paid at media as well.
- 15:57Uh, so that's full change.
- 15:58A lot of things when it comes to SEO, when, when it
- 16:01comes to how we do marketing.
- 16:03Reid Robinson: Yeah, that's coming.
- 16:05I mean, that's, you know, if you're familiar with the company called
- 16:07perplexity, uh, they announced last month, two months ago, something like that, uh,
- 16:12that they're, they're working on ads.
- 16:14Um, that they will have ads, um, in their side.
- 16:17By the way, I did, while we were chatting, test what are some tools
- 16:20marketers can use to automate their jobs.
- 16:21Zapier did come up, so, um, yeah.
- 16:23Actually,
- 16:24Itir Eraslan: I mean, before I, before our conversation this week,
- 16:27I was doing a lot of search about automation, marketing automation.
- 16:30Zapier comes a lot, and actually your website comes a lot, because of the
- 16:34fact probably that the, you know, burning questions that a marketer
- 16:39would have around automation, you answered them quite well at the website.
- 16:43So my other question would be like, um, so I'm a marketer of a medium sized brands.
- 16:50I don't have too many resources because, you know, we know that at
- 16:53big companies, there are teams that are adopting, testing these tools and
- 16:58adopting it to, and cascading down these tools for the use of automation.
- 17:02other employees.
- 17:03So, you know, when you are a part of a corporate company, big company, somebody
- 17:07will, you know, make the limitations for you and then we'll say, okay, we were
- 17:12going to use this automation platform, but in a medium or small size brands,
- 17:17there's probably one marketer that's sitting with a very small budget.
- 17:21And then everyone is talking about automation.
- 17:24And then you open web and you see like thousands of apps,
- 17:29thousands of people talking about.
- 17:31the new AI models.
- 17:33So where can I start as a medium size brand that I have no clue where to start?
- 17:39It's everyone is using chat GPT and then you pay the subscription for chat GPT.
- 17:44But what else?
- 17:45Because we know that AI and automation is more than what's
- 17:48chat GPT offers or Gemini offers.
- 17:52Reid Robinson: Yeah, that's a fun one.
- 17:54So I think to start, I think one of the things I always try to help
- 17:57clarify is a difference between.
- 18:00using chat to PT, like the application that you're probably familiar with.
- 18:03With like, I think they actually did buy the domain now for chat to PT.
- 18:06com versus what I would consider using in your sleep.
- 18:13Right.
- 18:13And that's kind of the world of chat to T plus Zapier, uh, to give an example.
- 18:17And the, the way that that works is through open AI APIs.
- 18:21This gets slightly complicated because.
- 18:25They're, they're separate, like OpenAI's API access and ChatGPT, the product,
- 18:29are two separate things, um, first of all, but an example of the differences
- 18:35with ChatGPT, if you're trying to personalize an email campaign, right,
- 18:40you would be, I don't know how you would do that manually, but I guess you would
- 18:43like Get some insights from the, a new lead that came into your email, right?
- 18:49Like you got a Facebook lead ad that came in, they gave you some information.
- 18:52You could take that, go to chatjpd.
- 18:55com, enter it, tell it some information about your business and
- 18:58ask for a good intro email, right?
- 19:01Or a good first touch email.
- 19:02And.
- 19:04Then copy that over to Gmail and send it with tools like Zapier, you can
- 19:09automate that whole process, right?
- 19:11You can actually now just say, Hey, every time I get a new thing from a
- 19:14Facebook data, send some information to chat to TV, uh, like an API call, um,
- 19:20and ask it the same sort of question.
- 19:23And then whatever it replies, put into either send that email right away if
- 19:28you're trusted or put it into a draft step which you could allow you to have
- 19:33like a manual step for you to review that content before it goes out.
- 19:37That's a key differentiation.
- 19:39I want to be clear to people.
- 19:39I think I talked to people who are like early on who are struggling to understand
- 19:44the opportunities that are available when they start to kind of put this like under
- 19:48the hood or run while you sleep or any way that helps you think through that.
- 19:52That's number one.
- 19:54Number two, go back to what I said before about like, you know, you mentioned like
- 19:57you saw Zapier a lot because of like our great content, that that's usually a
- 20:01critical thing and a huge bottleneck for companies, particularly midsize companies.
- 20:06So I would think like, how can you use it somewhere to either help you
- 20:09repurpose your content, accelerate that process, help you review it, you
- 20:14know, all sorts of things, but also just an internal process too, there
- 20:18is so much internal process within.
- 20:20Companies, especially mid sized companies and larger companies,
- 20:23um, that is very tedious.
- 20:26So I'd say one is, you know, thinking of those times where you have to write up a
- 20:31proposal for a piece of content, right?
- 20:35Um, you might just have a general gist of what you want.
- 20:39And you then are sitting in front of a form with maybe like 25 different
- 20:43fields that you need to fill in.
- 20:46That's a great opportunity for like a synchronous use case, or like a
- 20:49real time use case, where I would potentially take the form, copy and
- 20:53paste it over to chatGPT, and tell it the gist of what I'm hoping for a piece
- 20:57of content to do, who I'm targeting.
- 21:00What the objective is, right, like the information at a very
- 21:03high level that I understand, have it spit out responses for that.
- 21:09So then I can say like, okay, that was good.
- 21:11That was bad.
- 21:12Oh, I need to tweak this a bit.
- 21:13Those things that will help you take, like, Minutes off the task that you're
- 21:18doing are, I think, great opportunities to try to get familiar with the technology
- 21:23as well, because then you start to say like, Oh, you know what, we have
- 21:27the same thing where things come into our Trello board, but then we need to
- 21:31convert them to Airtable records, right?
- 21:33Like, but they're not the same format.
- 21:36The great opportunity to then be like, Oh, well, can I use an AI step here to say,
- 21:40Hey, these are marketing content requests.
- 21:43They come to Trello from our sales team because sales uses Trello, but
- 21:47in marketing land, we use Airtable and Hey, can you help us make, like,
- 21:52get these from one spot to the other.
- 21:54and
- 21:54Itir Eraslan: you, you can all do at ChatGPT, like these integrations
- 21:58from Trello and to the Airtable integration, like combining these
- 22:02two on top or something else.
- 22:05Thanks.
- 22:06Reid Robinson: So like that would in the world of Zapier land, the way
- 22:09that Zapier works is this concept of like triggers and actions.
- 22:13So triggers are like anytime something happens somewhere and actions
- 22:16are then what you do with that.
- 22:18So, you know, examples of like triggers would be like, every time I get a
- 22:21new email, a new message in slack, a new, uh, item or new row added to an
- 22:28air table database, things like that.
- 22:30Um, and then actions would be like, Um, uh, sending emails, creating
- 22:33drafts, adding rows to an air table, adding Slack channel messages,
- 22:38um, all sorts of things like that.
- 22:40So Slack bots, again, you know, I'll pick on content requests again.
- 22:44If you're a content request team, you could build a little Slack bot through
- 22:47Zapier that, um, When people message your team in Slack, maybe you want to
- 22:51say like, Hey, if this message is for content requests, can you reply to the
- 22:55user along with like our, I don't know, tips for writing content or tips for
- 23:00getting requests to help them formulate their request to our team better?
- 23:06And then it can automatically reply to that user or that
- 23:10employee in Slack, right?
- 23:11Like you can have these AI steps used in Slack type workflows too,
- 23:15which really help internally.
- 23:16Right.
- 23:18Itir Eraslan: But for example, when you, you work with a variety of companies
- 23:21from big companies to smaller scale companies, as far as I understand, in a
- 23:27smaller scale or a medium sized brand, for example, company that comes in, who
- 23:32is the owner of that integration, uh, that you deal with, who is making sure that
- 23:38this is adopted, From a variety of teams, because, you know, there's like usually
- 23:44one marketing person, one salesperson, or like a sales team of sometimes 30 people.
- 23:50And then there's only one person in marketing who's trying to, you know,
- 23:54get the requests of all the sales team.
- 23:56Who's that person?
- 23:57Who is it like the IT or is it the marketer or.
- 24:02Yeah.
- 24:03Yeah.
- 24:03Chief Technology Officer, if there is,
- 24:05Reid Robinson: especially for medium sized businesses, it's gonna be a
- 24:07marketing ops person a lot of the time.
- 24:09Mm-Hmm.
- 24:10. Um, first of all, I'd say shout out to all the team of one marketers out there.
- 24:13Um, if it, if you're listening, it's you, you probably knew that already, that,
- 24:18that's quite honestly mostly the answer.
- 24:20Uh, we see from that size of a business.
- 24:22Right.
- 24:23We do see like.
- 24:24You know, we could go all the way up to larger enterprise businesses, and for
- 24:27that, you do get into deployments that are like owned by it and then enabled
- 24:31across the organization, but I would say the ways happier typically starts in the
- 24:35business and grows is through departments and through the individuals, particularly
- 24:41folks like marketing ops, recruiting ops, any of those sorts of things where they're
- 24:46the folks who already understand how.
- 24:50Like those workflows, uh, interact, and now they're learning that they can
- 24:54actually like automate some of those workflows or make them more efficient
- 24:57across their organization, uh, starts to open up a lot of possibilities.
- 25:01Itir Eraslan: What are the things that they come up with
- 25:04when they request an automation?
- 25:06They'll probably have an idea.
- 25:08Okay, I want to automate content or I want to automate this or that.
- 25:13What, what's that thing that they usually come up with?
- 25:16The first
- 25:17Reid Robinson: I don't necessarily know because that is
- 25:19in terms of use cases broadly.
- 25:22If we're speaking more broad than AI lead management, lead capture,
- 25:26lead generation and lead management, like anything related to leads.
- 25:29It's very big, which makes sense, um, particularly because as anybody
- 25:33in that space knows, response time is like absolutely critical, right?
- 25:38So if you're generating new leads, particularly from an ad
- 25:40source like Google ads or LinkedIn ads or Facebook ads, right?
- 25:43Anywhere like that, where someone is like giving you their information and they
- 25:46want to be contacted about something, you don't just want to be like, all right,
- 25:50well, let's get back to them in like 48 hours when someone can call them.
- 25:55So that becomes a big thing with Zapier because the way that folks want to like
- 25:59Get that information into, you know, the thousands of serum tools that exist in
- 26:03the world, the spreadsheet tools they might want to reach out right away.
- 26:08That's again where like AI steps we've seen, uh, be used where they're
- 26:12personalizing that first touch response.
- 26:14So it's not just the same email every single time.
- 26:18Yeah, so that's one of the strongest ones that we see a lot
- 26:22of, particularly from marketers.
- 26:25Itir Eraslan: So, um, which of the functional areas?
- 26:29So we've talked a lot about how AI would help a marketer to automate things.
- 26:33Uh, but what are some areas that it's going to be?
- 26:37Slower to adopt because of the habits and because of the learnings that we
- 26:41have from back, that's a human flaw.
- 26:44Uh, but there is also the other part of AI, which the technology will
- 26:48be a little bit harder to progress.
- 26:51So what do you see there?
- 26:53Reid Robinson: I think the first thing I'll point out is really
- 26:56interesting, at least internal data.
- 26:57We don't, I don't have a lot more broader data to share, but the,
- 27:01one of the interesting trends we've seen from us in our internal
- 27:03adoptants for reference for anybody, we're about an 800 person company.
- 27:07And AI adoption for us is, I think we've got, I forget the exact stat,
- 27:12but around like 30, 35 percent of all employees are using AI automation today.
- 27:18And that breakdown is almost identically proportional across the entire company.
- 27:24Every team seems to adopt it to the same degree as every other team.
- 27:29Um, which is fascinating.
- 27:30We did not expect that result.
- 27:33And this is us with our own company, right?
- 27:35We didn't expect that result.
- 27:37So.
- 27:38I'm very curious to see, uh, as more and more data comes out in the world.
- 27:42How that looks for other businesses.
- 27:45I do think there are jobs that are certainly harder, uh,
- 27:48for these models to do today.
- 27:49Not like numbers is the biggest one, right?
- 27:52When
- 27:52Itir Eraslan: you mean numbers, what do you mean exactly?
- 27:55Reid Robinson: Like dealing with, dealing with numbers.
- 27:57Um, like these are still language models.
- 27:59Um, OpenAI does have things like the data analysis and stuff, but,
- 28:03uh, accuracy is still not there to really rely on these functions.
- 28:08Um, and so I think that's an area that becomes very difficult, uh, for them.
- 28:14Um, particularly if you're, you know, as you get to be a larger and larger business
- 28:17and want to rely or need to rely on that.
- 28:20Um, yeah, that'll be probably one of the slower ones to adopt.
- 28:24Uh, is anywhere where you, you're trying to use it over like data
- 28:27sets of numbers, difficult today.
- 28:31Um, getting better in the last year, it's been incredible improvements, but
- 28:34I think that'll be a much harder one than some of the more, uh, I don't know,
- 28:38freeform text type of things there.
- 28:41Other, in terms of team side, yeah, I don't know, because every team
- 28:45has their own, like IT for instance, um, you know, people may be like,
- 28:49what is an IT team going to do?
- 28:51But a big use case that I had come up with a lot from folks in IT, Was, you
- 28:56know, Hey, we get a lot of messages from employees in Slack complaining
- 29:01or not complaining, but like bringing up issues with their laptops or
- 29:04with software they're running.
- 29:05Right.
- 29:05But those teams need to get those requests into an actual queue, right?
- 29:10Like an actual request management tool, like Jira or something.
- 29:14Um, and again, as I was saying before, they, those teams were spending, uh, had
- 29:19one person do the math on this and it was an incredible amount of time that,
- 29:22because they would have to take that.
- 29:24The freeform text and put it into a structured ticket where it needed like,
- 29:28okay, who brought this request up?
- 29:30What was the issue?
- 29:31Who should this go to?
- 29:32Who should be triaged this likely, right?
- 29:34Like there was a lot of fields that needed to be filled in.
- 29:37And also
- 29:37Itir Eraslan: categorization involved.
- 29:39Exactly.
- 29:40Reid Robinson: And that was a step where they were able to automate, I think with
- 29:43almost 90 percent accuracy through Zapier with like the AI steps in the middle.
- 29:48Because now they were just like, hey, like you need to extract this data from these
- 29:52messages and like put it into this queue.
- 29:55And they saved like hundreds of hours.
- 29:58Um, and especially it was like menial hours where they were like just copying
- 30:01and pasting these values from, um, So anyway, just pointing out that there
- 30:07are these use cases that I think are relevant to all sorts of functions.
- 30:13Um, yeah, I think any, any time you're trying to deal with Data
- 30:15and numbers is probably going to be the slower things to see adopted.
- 30:20And of course, like things like sensitive information and whatnot.
- 30:22I won't, I won't get into that whole side of things with it right now.
- 30:25Itir Eraslan: Cultural, uh, you know, security.
- 30:29Yeah.
- 30:29Yes.
- 30:29Reid Robinson: Yeah.
- 30:30There's, I would say the more obvious things on that side.
- 30:33But yeah, less obvious.
- 30:34Yeah.
- 30:34Data is probably the one I would say.
- 30:37Itir Eraslan: One question.
- 30:38When you start working on projects, because you happen to work with a lot
- 30:42of different types of clients, uh, where some of them are providing much
- 30:47more information, some of them are not.
- 30:49So when you start a project, what, which projects flow.
- 30:54Easier, smoother, and faster.
- 30:56Are there some things that help you go and automate, help them automate?
- 31:02Reid Robinson: Well, number one broadly, before AI, is
- 31:05familiarity with their systems.
- 31:07Right, like if someone's like, oh, I want to automate something with Marketo.
- 31:09Right?
- 31:10I'm like, okay, do you know your Marketo system?
- 31:13If the answer is no, we're going to have problems, right?
- 31:17Like, that's a bit of an issue.
- 31:19Um, or if you're like, I want to deal with extracting this information from HubSpot.
- 31:23I'm like, okay, do you know, like, what field you're looking for in
- 31:27HubSpot and what object that's on?
- 31:29If you don't know that, trying to automate it is going to be difficult.
- 31:32Right?
- 31:33Um, so just knowing that, like the source of the destination of your information
- 31:38within AI, the thing that becomes critical, I would say, is getting these
- 31:42models context, uh, typically in the workflows that are valuable, you're not
- 31:47running things where you're just blanket asking the model to give you something.
- 31:51You know, you're not just saying generate a blog article about automation.
- 31:54Right?
- 31:55Like that's not going to get you a high value output.
- 31:58More likely, what you want to ask is, we're this type of business, with
- 32:02this audience, with this focus area, there's, you know, like mad libs, right?
- 32:07Like you want to do an advanced mad libs.
- 32:09And the challenge becomes figuring out where in your organization and where in
- 32:16the world, if it's the wider web, you can get the, Fill in the blanks, right?
- 32:22Like if you're trying to do personalized email campaign, do you know where in your
- 32:27CRM you have information on, you know, things like that would be helpful, like
- 32:32job title, uh, what products they use?
- 32:35Do you have anything record about their use case, right?
- 32:38Like what can give you information that when combined and provided
- 32:42to like an LLM will give you an output that feels valuable.
- 32:47And typically that's going to come with context.
- 32:49Uh, but actually.
- 32:51Being clear on where and how to get that context and those fill in the blanks, uh,
- 32:56becomes a challenge, but when I've worked with customers who do come with, you
- 33:01know, Hey, I'm familiar with our systems.
- 33:03We have this idea in mind and typically a good way to experiment with this,
- 33:06by the way, is like, um, is with chat to BT or with the open AI playground.
- 33:10If you like, can just.
- 33:12Right, what you're thinking about asking it and throw in the context
- 33:15of like a sample record, right?
- 33:17Like it's maybe real data to start to get some confidence of like if it's possible
- 33:22You can just like try to test around with like, okay when we gave it Job title.
- 33:27It gave better responses when we gave it, um, that customer's list of existing
- 33:32products, it gave better responses, right?
- 33:34Try to get some semblance of what gets you a valuable response,
- 33:41depending on what you're trying to do.
- 33:43Itir Eraslan: Perfect.
- 33:44So, um, back to the same question of.
- 33:47Okay, I'm a marketer in a medium sized brand and I don't have too much budget.
- 33:53I only use chat GPT to, you know, produce some social media content and so on.
- 34:00What would be the next step?
- 34:01Is it going to be like, Starting to work with, uh, a company like software
- 34:07automation company like Zepp here.
- 34:09What would be the next step for me, for example, if I'm a marketer just
- 34:12using ChatGPT to generate posts, to generate blog posts, and that's it?
- 34:18Reid Robinson: I will say there's interesting things, like,
- 34:20especially budget constraint.
- 34:21Uh, Google Gemini, for instance, uh, has a very generous free tier.
- 34:27Uh, for their API usage.
- 34:29So I'd highly, highly recommend checking that out.
- 34:31I think it's something like, what is it, a thousand requests a day?
- 34:35Um, which is, it's just a lot.
- 34:38Requesting like jobs, tasks, whatever word you want to think about.
- 34:41Like times you can interact with them.
- 34:43The model in a workflow.
- 34:45Um, so that'd probably be number one, um, on the budget side.
- 34:50Number two, I would say try to be T is a great, like, especially they now they
- 34:55opened up the free tier a little bit more.
- 34:58Itir Eraslan: I
- 34:58Reid Robinson: would say, as I said before, try to figure out what your
- 35:02life prompt is that you think if you were to be able to automate,
- 35:06this would get you valuable.
- 35:08Stuff, right?
- 35:10Uh, I wouldn't recommend testing your prompts once you've got
- 35:13your automations running.
- 35:15You kind of want to have confidence of, like I said, Hey, I found if I asked
- 35:21this type of question with this type of context, I get a valuable result.
- 35:25And if I could actually.
- 35:27Automate that and have it run while I'm sleeping.
- 35:31Fantastic.
- 35:32Like that's going to be really valuable for the business.
- 35:34Um, I think experimenting in that way can really help you because then you're
- 35:39not building a whole workflow and whole system and then trying to get, because
- 35:45that's a very difficult part to get right.
- 35:47I'll be clear.
- 35:48Um, and it's also the part that like.
- 35:51Once you can get right, I think makes it, it kind of illuminates, Oh, wow.
- 35:56It was really good at doing, um, I don't know, like recommending the intro
- 36:02paragraph of an email, but it didn't do the rest of the body really well.
- 36:06Like maybe you could only get good confidence on the intro paragraph.
- 36:09Well, now, you know, like you want to build a workflow that's going to get you.
- 36:13That intro paragraph, and maybe you figured out like what sort
- 36:16of context fields are needed.
- 36:18So that would be the other thing I would recommend is, and also
- 36:21the base models are, you can just talk to them about it too.
- 36:25They do seem to have a lot of information on Zapier and other tools.
- 36:29So just ask it as well.
- 36:31Be like, Hey, I'm a marketer for this business and I use these tools, right?
- 36:36Recommends things that I can do.
- 36:38That was why I built like our little automation consultant GPT,
- 36:41uh, that anybody can use as well.
- 36:43Yeah, some funky things you can start to do to kind of, again, that's what I
- 36:46meant by like, you kind of stretch your capabilities because all of a sudden
- 36:50you might have access to a wealth of knowledge of automation and marketing.
- 36:57Itir Eraslan: Yeah.
- 36:57Yeah.
- 36:58I mean, I always ask like day to day, sometimes like, especially
- 37:01before podcasts, I asked chat GPT.
- 37:03Okay.
- 37:04So what areas in marketing, like the questions that I'm
- 37:06asking to you, I ask it.
- 37:08So see what they come up with.
- 37:10Uh, and for example, the thing that you said about numbers.
- 37:14Uh, that is like, it's a little bit tricky.
- 37:17It never came up, although I was asking.
- 37:20Similar questions.
- 37:21So that's why, I mean, I always believe that still human interactions
- 37:26are quite valuable in business.
- 37:27And also these models, as I understand, it's briefing.
- 37:32The model is one of the most important things, just like working with an
- 37:36artist or a creative person, the way, how you brief that person is so
- 37:41important that if I brief that person, or if someone else briefs you're going
- 37:47to get two different creative work.
- 37:50And so that's the value that we add as PSPRO.
- 37:53That's why I believe that who facilitates is as important as the model itself.
- 37:59Uh, although I, I'm really positive about like all this, um, you know,
- 38:03technological advancements of chat GPT or, sorry, that I keep on, you know,
- 38:07promoting chat GPT, but generative AI in general and AI going forward.
- 38:12But yeah, I also believe in the, who, Facilitates that conversation and
- 38:16that work actually, my last point would be around social media because
- 38:21you have, you are a student of social media because you have been in the
- 38:25social media world for so long.
- 38:27How do you think that all this automation, how do you think
- 38:31things will evolve in social media?
- 38:33Like, will there be any change on how we communicate or how we
- 38:36are going to use social media?
- 38:38Reid Robinson: Hmm.
- 38:39Hmm.
- 38:42I mean, it clearly is.
- 38:43I think the big thing that a lot of people are trying to do now is like automate.
- 38:46I mean, Usage of AI and social media has been around for a longer than
- 38:50I've been working in it since 2014.
- 38:53I want to say, um, right?
- 38:55Like no, anybody here is quite familiar with like bots on
- 38:57Twitter and stuff like that.
- 38:58Right?
- 38:59Yeah.
- 38:59Um, I think we see a lot of people now trying to like automate their like
- 39:04LinkedIn comments and stuff like that.
- 39:06Not the greatest use cases.
- 39:08I hope we see better use cases over time.
- 39:11Um, Yeah, like I said, I think for me right now, what I'm
- 39:15seeing, it's like converting and being able to repurpose content.
- 39:19Um, video, right?
- 39:21There's a big one.
- 39:21There's a lot of tools right now that are trying to help you take existing
- 39:23video assets and use AI to splice them into TikTok formats, right?
- 39:28It's why you see a lot of TikTok videos these days that feel
- 39:30a bit like choppy and split.
- 39:34Yeah, it's, and it's cool, right?
- 39:36Again, I recognize, you know, a lot of people are team of one on their
- 39:39efforts or they're being asked to do things that like, they're like,
- 39:42I never created a video before.
- 39:44Like, I don't know how to do that.
- 39:46And so yeah, if there's a tool that can get you something that
- 39:48feels like 80 percent good, you know, probably gonna use it.
- 39:53Um, so I think we'll see.
- 39:55More and more of that, I think we'll see like AI do a lot more in
- 39:58like ad creation, ad delivery as well from the platforms themselves.
- 40:02Um, I think that'll be a pretty cool part is when that kind of like
- 40:05model runs happen on the user level.
- 40:09Itir Eraslan: Yesterday I was commenting on a post of a friend, uh,
- 40:13and then, you know, I started trying a premium LinkedIn premium once in
- 40:17a while and then I just give up and then it, it's now recommending me.
- 40:22ai, uh, yeah, generated, uh, comments on posts.
- 40:26Of course I never use 'em 'cause they're so generic.
- 40:29And then I wrote a comment and that day I was using a lot of, uh, giants
- 40:33of ai and then I, at the end of the comment I said, I'm in like my.
- 40:38Posts and my comments sounds really like AI and I'm afraid because it's so generic
- 40:44And there was like a laughing moment, yeah
- 40:50Reid Robinson: Yep, it's a weird
- 40:51Itir Eraslan: but I feel like on my side that I feel social media is for The main
- 40:56purpose of social media is to connect people and then for people to communicate.
- 41:00And when I see automated comments on my posts, I just don't want to interact
- 41:07because I know that it's automated.
- 41:08I know that it's generated by a bot or whatever.
- 41:11And then it gets me, I'm quite bored.
- 41:14And whenever I see like a real comment from a person, then I, You
- 41:18know, interact and I want to use it.
- 41:21So in that case, I see an opportunity here because as a brand, if you don't
- 41:25overuse AI and then you try to make things different than how AI platforms are doing,
- 41:31then you have the opportunity to stand out in the crowd because everyone is, you
- 41:35know, It's probably adopting the same, same automated models and same automated
- 41:39formats, like, you know, TikTok formats that everything is very clear, same
- 41:44captions, same graphic design or so on.
- 41:47So that's how, I mean, although I'm a big believer of AI in terms of
- 41:51facilitation and in terms of helping out in projects, when it comes to tools.
- 41:56Too much of doing that, too much of automation, too
- 41:58much of, uh, relying on it.
- 42:01Uh, I think it's gonna, you know, create a lot of, uh, mundane and boring stuff,
- 42:07uh, that the companies that are not using it too much or just using it with
- 42:13this in mind, then they'll stand out.
- 42:16That's my personal take on the social media thing.
- 42:19Reid Robinson: Yeah, I would agree.
- 42:21That's why I came in and work in AI.
- 42:22I don't use AI to actually help you write LinkedIn posts or anything like that.
- 42:26I find it just, uh, sounds way too, I don't know, AIE is the
- 42:31best way to put it these days.
- 42:33I mean, your
- 42:34Itir Eraslan: posts are in a way quite, quite, It doesn't sound like AI.
- 42:37And then although you, your profile says you are building AI
- 42:41products, so this already gives a clue about where we are going.
- 42:45Reid Robinson: Yeah.
- 42:46I do use it sometimes to like help with, um, ideas for it, which is, I think
- 42:50like that ideation stage is fascinating.
- 42:53Um, and as well, I like getting it to write something and then
- 42:56like critiquing it into oblivion.
- 42:58As is maybe a good way to put it, where you're like, Hey, can
- 43:00you help me do a post about this?
- 43:03And then you see what it does, and you're like, Oh my god, what a horrible post.
- 43:06Let me write it.
- 43:08But it's like, I don't know what it is, but like that act of, It sounds horrible
- 43:12to say this out loud, but I guess that act of judging its output kind of gives
- 43:16me a better path to forming my own ideas.
- 43:21Um, because I'm like, oh, it said that very poorly.
- 43:23And then I'm like, okay, well, what would be a better way to say that?
- 43:26Um, and then it kind of helps me get going.
- 43:29So sometimes I can find myself get unstuck by getting a horrible output and then at
- 43:33least in motivating me to do a better job.
- 43:36Itir Eraslan: Perfect.
- 43:37Reid Robinson: Yeah.
- 43:37Itir Eraslan: So two personal questions.
- 43:40I'm not personal, but that I asked to every other guest.
- 43:44Uh, if you could recommend one book, either it's be a business book or
- 43:49a regular book, what would it be?
- 43:51Reid Robinson: Yeah, I mean right now I'm extremely into the Red Rising series,
- 43:54uh, it's a sci fi book, uh, absolutely love it, frankly, I like Hunger Games.
- 43:58I've never heard about
- 43:58Itir Eraslan: that, by the way.
- 44:00Reid Robinson: Ah, well if you like Hunger Games, it's kind of similar
- 44:04type of feel, if you will, it's a bit darker, um, honestly, but really good.
- 44:10Um, however, more I don't know, relevant.
- 44:13I really liked this book by Elizabeth Gilbert, who is the
- 44:16author of Eat, Pray, Love.
- 44:19She wrote a book called Big Magic, which is all about creative
- 44:22thinking, um, and things like that.
- 44:25Um, which I actually found very helpful for myself.
- 44:28Um, one of the things that I often get involved with is kind of like
- 44:33ideation and that from like both a product and use case side of things.
- 44:39Sometimes it can be really challenging to like think through
- 44:43these, I don't know, interesting ideas with new technology, right?
- 44:47And it was fascinating to me to read a book from an author focusing on this
- 44:52concept of creativity and fostering like creative endeavors and thinking.
- 44:57Which was like, she was very focused for I think the readers who are mostly
- 45:00people who wanted to write books.
- 45:02But it was cool because you know, it's like, Something like I would absolutely
- 45:06have never, like never think about writing a book, but it was cool to think
- 45:10about how, um, she thought about her creative process and to think through
- 45:14the parallels of how that relates to, you know, even something as mundane as like
- 45:20the creative process of a LinkedIn post.
- 45:22Right.
- 45:23And like, what might be a relevant and fun thing to think about
- 45:27and ideas and even use cases.
- 45:29Like I.
- 45:30Um, I'll give you one really, she had a very interesting concept of like ideas
- 45:35for books was her thing and how they like, she had this concept that ideas
- 45:38kind of float around the world and they just like bounce into random people and
- 45:43stick in their heads and if you don't, you know, it might hit someone else as well.
- 45:47Um, and the interesting thing for me, though, was this concept of.
- 45:52Acknowledging the idea and processing it and being comfortable
- 45:56to say no to the idea and move on.
- 45:59Um, cause I think one thing that's challenging with a lot of the
- 46:01tools that are out there and just the pace of things too, it's like,
- 46:04Oh, I want to, I want to try this.
- 46:06I have this idea for this.
- 46:07And then you might need to say like, all right, you know what?
- 46:09It's a cool idea, but this idea is not for me.
- 46:12Like, thanks for coming to me idea.
- 46:15Go have a great day.
- 46:16I hope you find someone like cool out there to work with you.
- 46:19Um, really?
- 46:20I mean, that's,
- 46:21Itir Eraslan: that's the book that I've read a while ago.
- 46:24It's been a few years.
- 46:26I think she published that book, but I didn't read her main
- 46:29book, which is Eat, Pray, Love.
- 46:30I, instead of that, I read the Big Magic.
- 46:33It's a pretty nice one.
- 46:35Um, so what's your favorite cafe place in Vancouver?
- 46:39Reid Robinson: Ooh, if you're in Vancouver, I really like Milano coffee.
- 46:43Itir Eraslan: Milano.
- 46:44Reid Robinson: Yeah, there was one near the Hootswood office
- 46:45in the area of Mount Pleasant.
- 46:47They had great coffee.
- 46:48There was one spot in particular that was over a park that
- 46:50looked over the mountains.
- 46:51Very pretty.
- 46:52Uh, if you ever, for anybody who's been to Vancouver, it's like
- 46:55number one thing people tell me.
- 46:56It's like, Oh my
- 46:56Itir Eraslan: God,
- 46:56Reid Robinson: it's so
- 46:56Itir Eraslan: pretty there.
- 46:57Reid Robinson: Um, so
- 46:59Itir Eraslan: that's what they say.
- 47:00Yeah.
- 47:01Yeah.
- 47:02A lot of my friends come over there to, to ski and snowboards.
- 47:05Um, but I've never been, then I hope to be there soon.
- 47:10Uh, Reid, thanks a lot for your time on a midday and midweek.
- 47:16Uh, I really appreciate it.
- 47:17And it was nice to meeting you as well.
- 47:19Thanks for being a guest on my podcast.
- 47:22Reid Robinson: Thank you so much.